AI agents
Agents that plan on a model they can prove.
Rexfin's agents analyze, model, and orchestrate over one reconciled financial model, every answer grounded in figures your team can trace back to the filed page it came from.
Works with what you file
- PDF filings
- Bank statements
- ERP exports
- ZATCA e-invoices
- Excel schedules
- Board packs
- QuickBooks exports
- Xero exports
- NetSuite exports
- Sage Intacct exports
- CSV ledgers
- Snowflake tables
How the agents work together
Decisions get better when every agent shares one ledger.
Each agent is scoped to one job and reads from the same reconciled model, so a question the Analyst answers and a scenario the Modeler builds never quietly disagree. A lightweight routing layer sends each request to the agent built for it; nothing routes around the reconciliation the model already enforces.
"We didn't want an agent that sounds confident. We wanted one that's checkable."
The Rexfin team
Agents overview
Meet the agents working your model.
Each agent is grounded in the same reconciled figures, so nothing it says or builds drifts from the ledger.
Ask anything; get an answer with its receipts.
The Analyst agent retrieves figures straight from your reconciled model, runs the calculation, and explains what moved: every claim carries the citation and computation it came from, not a guess dressed up as one.
- Data retrieval
- Variance narration
- Cited answers
Q3 revenue by segment
Gross margin +3.2pts on lower COGS p.24
Describe the plan; the model builds it.
Say what you want to change (a hiring freeze, a new price, a downturn case), and the Modeler agent edits the model inside your existing definitions, so the change lands the way finance actually models it.
- Scenario building
- Driver logic
- Governed definitions
Scenario · Hiring freeze
Explore every scenario before you commit to one.
In development: the Scenario agent will run scenarios across your model in parallel and surface the trade-offs, on the same trust chain as every other agent; nothing it shows skips reconciliation.
- Scenario comparison
- Trade-off surfacing
Scenario comparison
Base
Upside
Downturn
Give your own tools the same governed access.
The MCP server exposes the model under the same permission and citation rules as the product itself, so any AI tool or internal workflow your team already runs can query it as a fully governed agent, with no separate export and no untracked copy of the numbers.
See developer access ↓name: internal-fpa-bot
access: mcp-server
scope: read + cite
permissions: role.controller
on_missing_citation: refuseGrounded by design
Every agent answer sits on the same five-step chain.
- Filed
- Extracted
- Reconciled
- Cited
- Exported
Developer access
A programmatic door into the same governed model.
The Rexfin MCP server gives other AI tools and internal systems query access to your model, under the same permission and citation rules enforced everywhere else in the product.
It's the same reconciled figures, the same lineage, the same refusal to answer a number it can't trace, just reachable from wherever your team already builds.
Talk to us about MCP accessReady when you are
Put a governed agent behind every number.
Book a demo and ask the Analyst agent a real question about your own actuals, with the citation attached.